Skip to content
Fundament Docs

ADR-0002: Design scale target

Proposed
Status

proposed

Date

2026-03-09

Group

cross-cutting

Depends-on

ADR-0001

Context

Fundament must define its design scale to make informed architecture decisions. Components that work at 500 physical servers may not work at 50,000. The target determines which tools are viable and which operational models are required.

Options

Option 1: Single-org scale (up to 500 physical servers)

  • Pros: most tools work out of the box; small team viable; fastest time to production

  • Cons: limits the platform to a single organization or small consortium

Option 2: Medium scale (up to 5,000 physical servers)

  • Pros: broad tool support; proven architectures available; serves multiple organizations

  • Cons: may require re-architecture if demand grows beyond this target

Option 3: Hyperscale (up to 50,000 physical servers / ~2M logical CPUs)

  • Pros: meets stated government ambition (30-50% sovereign compute); future-proof for multi-year growth; serves all government organizations on one platform

  • Cons: eliminates many tools that are not proven at this scale; requires automation-first approach from day one; dedicated DC network automation required

Decision

Design for 50,000 physical servers / ~2,000,000 logical CPUs, to be reached within 3 years. Options 1 and 2 were considered but do not match the government ambition to consolidate 30-50% of sovereign compute onto a single platform. Designing for a smaller target risks costly re-architecture when demand exceeds the initial design. All architecture decisions must be validated against this target.

Consequences

  • Tools must be proven or credibly validated at this scale

  • Manual operations are not viable — everything must be automated

  • DC network automation is a first-class concern

  • Capacity management and monitoring must scale horizontally